基于实值遗传算法的模糊神经网络辨识器

提出了一种基于实值遗传算法(RVGA)的模糊神经网络辨识器·它常被用于非线性动态系统的辨识·通常模糊神经网络辨识器参数的训练采用反向传播学习算法(BP),但是用BP算法有训练时间长,容易陷入局部极小的问题·采用RVGA来训练模糊辨识器的参数,由于GA算法具有并行运算,多点寻优等特点,所以它运算速度快,容易实现全局寻优·传统的GA算法采用二进制编码,计算繁复且占用大量的空间·采用一种新的实数编码方法,在实数域上进行遗传运算,操作简便,特别适用于需要调整的参数较多的情况·仿真结果表明,该辨识器具有良好的逼近性能和较快的训练速度·

FNN Identifier Based on Real-Valued Genetic Algorithms

A kind of fuzzy neural network (FNN) identifier based on real valued genetic algorithms (RVGA) was proposed. It is often used in the identification of the non linear dynamic systems. The parameters of the FNN are usually trained by BP algorithm. But BP algorithm has some disadvantages, such as long training time and the falling of the algorithm into the regional smallest points easily. RVGA was used to train the parameters, and a kind of real code was adopted to represent the chromosome. The GA operations are operating in real domain. The advantages of the identifier were showed by the results of simulation.